SKEPTICISM AND VARIETIES OF TRANSCENDENTAL ARGUMENT

Transcription

1 SKEPTICISM AND VARIETIES OF TRANSCENDENTAL ARGUMENT Hamid VAHID ABSTRACT: Transcendental arguments have been described as disclosing the necessary conditions of the possibility of phenomena as diverse as experience, self-knowledge and language. Although many theorists saw them as powerful means to combat varieties of skepticism, this optimism gradually waned as many such arguments turned out, on examination, to deliver much less than was originally thought. In this paper, I distinguish between two species of transcendental arguments claiming that they do not actually constitute distinct forms of reasoning by showing how they collapse into more familiar inferences. I then turn to the question of their epistemic potentials which I argue to be a function of both their types as well as their targets. Finally, these claims are reinforced by uncovering links between certain recent claims about the efficacy of transcendental arguments and the so-called Moore s paradox. KEYWORDS: transcendental arguments, skepticism, situated thought, Moore's paradox On the surface, at least, there does not seem to be much disagreement among philosophers about the salient features of what are known as Kantian transcendental arguments. These include their primary function (refuting varieties of skepticism), their subject-matter (experience and thought) and their structure (an extended modus ponens argument whose conditional premises are supposed to express necessary conditions for the obtaining of their subject matter). Kant s idea in proposing transcendental arguments was to identify the necessary conditions of the possibility of experience which he took to include the truth of the propositions that the skeptics generally seek to deny. Recent years have however witnessed major shifts of emphasis in regard to the nature of these features. Many theorists have cast doubt on the anti-skeptical potentials of transcendental arguments. Some have broadened their scope to include the possibility of language, self-knowledge and the like. And, as yet, there is no consensus about the status of the modality of their conditional steps. In this paper, I begin by distinguishing between two species of transcendental arguments claiming that they do not actually constitute distinct forms of reasoning by showing how they collapse into more familiar inferences. This would then pave LOGOS & EPISTEME, II, 3 (2011):

2 Hamid Vahid the way for examining their epistemic potentials which will be argued to be a function of both their type as well as their targets. Finally, these claims will be reinforced by uncovering surprising links between certain recent claims about the efficacy of transcendental arguments and the so-called Moore s paradox. 1. Varieties of Transcendental Argument As already noted, transcendental arguments are, formally speaking, extended modus ponens inferences with their conditional steps consisting of premises expressing necessary conditions for the obtaining of certain states such as experiences, thoughts, etc. p p q q r.. * t Here p is supposed to express certain facts that even the skeptic ought to accept while the conclusion t is something that he typically denies. Kant, it seems, thought that some of his arguments in the Critique which are of such a type establish and reveal certain truths about the world, e.g., that we can have experience only if reality is indeed causal, that perception of change is possible only if there are persisting substances and so on. A number of theorists have, however, found transitions from psychological premises in these arguments to conclusions about an objective, non psychological world puzzling and invalid. Barry Stroud, for example, has claimed that this subjective/objective gap can only be bridged with the help of some extra assumptions (like the verification principle) which would in turn render transcendental arguments redundant. 1 Stroud seems to think that the subjective/objective divide is something like the is/ought gap in ethics though he presents no general argument as to why closing the former is impossible. Thus, barring an argument, the possibility of such transitions remains wide open. 2 He does however muster some support for his claim by examining a number of (more recent) transcendental arguments arguing that their success does not really hang on the truth of their conclusions but only on our believing them, 1 Barry Stroud, "Transcendental Arguments," The Journal of Philosophy 65 (1968): For acknowledging this possibility, see, for example, Ralph Walker, "Induction and Transcendental Argument," in Transcendental Arguments, ed. Robert Stern (Oxford: Oxford University Press, 1999),

3 Skepticism and Varieties of Transcendental Argument and, indeed, Kant himself seems to have also suggested transcendental arguments that are merely intended to establish how and what we must think about the world. 3 Instead of getting bogged down in the debate over which type of such arguments deserves to be called transcendental, it would be more prudent to admit that there are indeed two species of transcendental arguments sharing the same surface structure, and then try to evaluate the epistemic significance of each species. We may thus say that just as there are objective and subjective readings of the formal axioms of the probability calculus, there are objective and subjective interpretations of the formal structure of transcendental arguments. Accordingly, while objective transcendental arguments seek to reveal truths about the world, the subjective variety is intended to show why certain beliefs are indispensable for having thoughts, experiences and so on. I call these first-order and second-order transcendental arguments respectively (reflecting the type of content their conclusions express). With this distinction in mind, I shall now proceed to show that, their surface structure notwithstanding, transcendental arguments are actually species of inference to the best explanation. As noted earlier, transcendental arguments purport to show that some conditions (C) are necessary for the obtaining of certain states or properties (E) (such as experiences of change, causality, etc.). However, as it turns out, it seems that all that such arguments do is to show that those conditions (together with certain background assumptions) are sufficient for the obtaining of E without being uniquely so. For example, in the Transcendental Deduction, Kant argues that for experience to be possible, it must belong to a unified consciousness which, in turn, requires that those experiences be experiences of physical objects. By itself, this does not rule out alternative ways of securing such experiences. Insisting on the impossibility of such alternatives seems to say more about the limitations of our imagination than those of the reality. 4 At most what the Kantian transcendental reasoning does is to narrow down alternative ways of securing such experiences by seeing, in Strawson s words, how [our hypotheses] stand up to attack. 5 Under these circumstances, it would be more reasonable to speak of the necessary conditions of experiences as being the most appropriate means (relative to a set of available, rather than possible, alternatives) of securing such states. 3 Walker, "Induction," See, for example, Stephen Korner, Fundamental Questions in Philosophy (London: Penguin, 1971) and Richard Rorty, "Transcendental Arguments, Self-reference, and Pragmatism," in Transcendental Arguments and Science, eds. Peter Bieri, Rolf-Peter Horstmann, Lorenz Kruger (Dordrecht: Reidel, 1979), Peter Strawson, The Bounds of Sense (London: Methuen, 1976),

4 Hamid Vahid To give a concrete example, consider Strawson s version of Kant s argument in the Deduction (often referred to as the objectivity argument ) which proceeds by showing why a particular alternative to the realist claim, viz., the sense-datum hypothesis, fails to satisfy what he takes to be the desiderata of experience. 6 This is done by trying to demonstrate which of these alternatives tell a better story about (or provide an explanation of) the obtaining of experience. This, in turn, involves examining which alternative hypothesis fits better with our (independently justified) background assumptions about the subject matter under discussion. Accordingly, Strawson proceeds by unveiling a number of (purportedly) justified claims about the nature of experience, e.g., that our experience possesses such features as unity, conceptualization, self-ascription and so on. He then considers two competing hypotheses, viz., the realist hypothesis emphasizing the existence of mindindependent objects and the sense-datum hypothesis construing these objects as bundles of sense data to see which alternative provides a better explanation of those particular features of experience. He concludes that the type of experience that is borne out by the sense-datum hypothesis fails to possess the kind of unity that our experience displays. Thus understood, we may see the function of transcendental reasoning as painting as coherent a picture or story of the obtaining of experience and its specific features: We have before us the materials of a transcendental drama In the Transcendental Deduction the story is told, the explanation is given. 7 This way of conceiving of transcendental arguments, however, encourages viewing them as species of inference to the best explanation (IBE) where a hypothesis is adopted from among a number of competing alternatives in virtue of providing the best explanation of the available data. 8 The so-called explanatory virtues that are said to guide our choice of the best hypotheses often include such properties as simplicity, parsimony, non-ad-hocness, etc. There is, however, a great debate as to how these notions should be unpacked. It has been more customary to view them as reflecting purely methodological and pragmatic principles implying nothing about the world. But this raises the worry as to why possession of such (pragmatic) virtues should make a hypothesis more likely to be true. A much more reasonable account of the nature of such virtues has been recently advanced by Sober who, rejecting the pragmatic approach, claims that appeals to such properties are actually surrogates for stating background assumptions about 6 Strawson, The Bounds of Sense. 7 Strawson, The Bounds of Sense, For an early suggestion along these lines see Jay F. Rosenberg, "Transcendental Arguments Revisited," The Journal of Philosophy 72, 18 (1975): , which views transcendental arguments as no different from a Peircean abduction (623) but does not substantiate it. 398

5 Skepticism and Varieties of Transcendental Argument the subject matter and the inference problem one faces. 9 For it is only relative to a set of background assumptions, he says, that observations can be said to support one hypothesis better than another (there is obviously nothing sacrosanct about these assumptions as they themselves may be challenged). So the function that properties such as simplicity are intended to serve is to help bring observations to bear on hypotheses. Viewed in this light, the credibility of an IBE turns out to depend on how well the selected hypothesis coheres with the rest of our (independently justified) background assumptions and this accords very well with our diagnosis of transcendental arguments. We may, thus, conclude that a Kant-style transcendental argument is actually an IBE in disguise. 10 By way of supporting this conclusion, consider a recent statement of some of Kantian transcendental arguments (due to Sacks 11 ). His first example is taken from the First Analogy where Kant seeks to show that all change is merely the alteration of an abiding substance. On Sacks reading, since our perception of change has a unified character, that is possible only if something persists across our experiences. This seems to be just another way of saying that the best way to account for (explain) the unified character of our relevant experiences is to postulate the existence of persisting objects. The second example is drawn form the Second Analogy where Kant tries to establish why everything that happens has a cause. According to Sacks, Kant s thought is that where one s perceptions are of an event, then the order of these perceptions is irreversible. The agent would then be faced with the puzzle of the irreversibility of two perceptions that can only be accounted for if the perceived states of affairs themselves happened one after another: [I]t s only thus that the subject can explain the irreversibility of his perceptions of them. 12 This is, again, another way of saying that the perceived irreversibility is best explained by postulating a corresponding relation within the objective domain Eliot Sober, Reconstructing the Past (Cambridge, Mass.: MIT Press, 1988). 10 For a fuller defense of this suggestion see Hamid Vahid, "The Nature and Significance of Transcendental Arguments," Kant-Studien, 93, 3 (2002): Here is a recent endorsement of this claim: Kant s most important method, the transcendental method, is also at the heart of contemporary cognitive science. To study mind, infer the conditions necessary for experience. Arguments having this structure are called transcendental arguments. Translated into contemporary terms, the core of this method is inference to the best explanation, the method of postulating unobservable mental mechanisms in order to explain observed behavior. (Andrew Brook, Kant s View of the Mind and Consciousness of the Self, The Stanford Encyclopedia of Philosophy, ed. Edward N Zalta (Summer 2007 Edition), 11 Mark Sacks, 2005, "The Nature of Transcendental Arguments," International Journal of Philosophical Studies 13, 4 (2005): Sacks, "The Nature," I have already emphasized that we must be content here with the best available explanations. 399

6 Hamid Vahid Before leaving this topic, it would be instructive to examine Sacks own account of the nature of transcendental arguments which promises to explain a number of peculiar features of such arguments. Although I am going to challenge his account of transcendental arguments, nothing in what he says contravenes the conclusions we have reached so far. Indeed, as we shall see later, his view of the matter provides an interesting gloss on my own account of the epistemic efficacy of transcendental arguments. Sacks main concern is with the status of the conditional or ampliative steps ( q r ) in transcendental reasonings. However, instead of taking their bare propositional content, q r, as constituting such steps, we should consider, he says, the situated thought of the content in question, viz., the thought that one would have from a particular point of view within a framework. For example, while the propositional content of q, say, there is a tree in the garden, doesn t entail r, there is a tree in front of me, the situated thought of q does. Consider an agent situated in front of a tree in a garden, thus, coming into perceptual contact with it. Under these circumstances, the agent can have the situated thought that there is a tree in the garden (s q ). Being in possession of s q, he can come to know, without further observation, that there is a tree in front of him. Sacks general idea is that while, to substantiate the conditional premises of a transcendental argument, an argument is needed to sanction the move from q to r, the situated thought of q would suffice to license a priori the ampliative move in question. For instance, while it is difficult to see why the central claim in the First Analogy, Every change is merely the alteration of an abiding substance, should be true, one can easily see how the move from the situated thought of something changing to the existence of a pertinent persisting object can make sense. The same goes for Kant s other claims like every event has a cause, and its accompanying transcendental argument. Here too, says sacks, the relevant situated thought would enable us to see why the claim in question is true. Sacks gloss on transcendental arguments raises a number of interesting questions. To begin with, the notion of the situated thought of a content does bear some resemblance to the phenomenalist analysis of the notion of an object (without its reductionist associations). Worried about the skeptical consequences of the realist claim that there is a genuine distinction between the conception we form on the basis of experience and the way things are, phenomenalists sought to close the appearance/reality gap by reducing statements about physical objects to those about our perceptions of them. Accordingly, a statement like q, There is a tree in the garden, is to be reduced to those describing our sensory impressions when we come into perceptual contact with the tree in question. This seems to be akin to what the thesis of situated thought involves. A situated thought of q requires that 400

7 Skepticism and Varieties of Transcendental Argument we. envisage a subject situated in front of a tree in the garden, and being perceptually related to it. 14 Moreover, just as phenomenalists qualified their initial thesis by admitting possible as well as actual perceptions in order to account for statements about unperceived physical objects, thus, analyzing, say, q in terms of such statements as if I were to see such a tree before me, then certain sensory impressions would follow, Sacks also makes a corresponding distinction between actual and possible situated thoughts: a situated thought differs from the corresponding experience in that the situated thought does not require that the subject actually be situated only that he approximate in thought to what would be delivered up to him if he were so situated. 15 An initial problem with the situated-thought thesis is that even if we can establish that there is a valid transition from the situated thought of the content expressed by q (s q ) to r, why should it follow that there is likewise a valid transition from q to r when it is, in fact, q r that appears as a premise in the pertinent transcendental argument? Another problem concerns the nature of a situated thought itself. Is it sufficient for a thought to be situated that an agent be perceptually related to the object of that thought? Sacks himself denies this by providing the following illustration. I am in a laboratory vehicle, receiving pictures on a screen from a camera that I know is in the garden, though I do not know that the vehicle is in the garden. This is not yet a case of a situated thought, for although my knowledge of q is based on my own perception, I fail to infer r from q. What is needed, he says, is that the thought in question be grasped from one s point of view and be informed by it. Only then the mere having of the thought licenses the move to there is a tree in front of me. 16 But it seems that these conditions still fail to be sufficient. To see this, consider the following example. 17 Suppose there is a candle in front of an agent but he doesn t see it directly. Rather, there is a series of elaborately disguised mirrors that reflect an exact image of the candle. These mirrors are placed next to one another in such a way that what the agent sees is the image of the candle in the last mirror in the series in front of him where the candle itself is and hiding it from his view. In this case, the agent s sensitive 18 and justified belief that he is looking at a candle is based on his own perception from which he can justifiably infer that 14 Sacks, "The Nature," Sacks, "The Nature," Sacks, "The Nature," Adapted from Gilbert Harman, Thought (Princeton: Princeton University Press, 1973). 18 This belief is sensitive in the sense that if there were no candles in the room the agent would not believe that he is looking at one. 401

8 Hamid Vahid There is a candle in front of me, thus, satisfying Sacks strictures on a situated thought, but, intuitively, we do not want to say that the agent s thought is really situated. 19 Let us now turn to the application of the situated-thought thesis to transcendental arguments. Sacks idea, we may recall, is that only by taking conditional sentences in such arguments as expressing situated thoughts can there be any hope of establishing their conclusions. He gives, as an example, Kant s claim in the First Analogy, namely, that All change is merely is the alteration of an abiding substance. It is difficult to see why this claim should be true, he says, but things will be different once we consider the situated thought involving its content. 20 But I do not see why we need to resort to the situated thoughts involving the changes, rather than the changes themselves, to substantiate the claim in question. Sacks himself provides a corresponding argument focusing not on the experiences but on the states of affairs apprehended: For [the state of affairs] to be presented as changing in relation to one another, or as abiding, there must also be an abiding external world on the backdrop of which it makes sense to say either that where one object was there is now another, or that where one object was there is still the same object. 21 It may be that Sacks thinks ascending to a situatedthought level would provide an immediate a priori route to the claim that there are abiding substances just as the mere having of the situated thought that there is a tree in the garden immediately licenses the move to there is a tree in front of me." But this does not seem to hold in the transcendental case discussed as Sacks provides an elaborate argument (filling a full paragraph) to show why the claim in question holds. Similar points apply to Sacks other example concerning causation. Finally, Sacks claims that what is both necessary and sufficient for a transcendental proof to work is only that the subject s thought be situated virtually. This claim seems to limit the scope of transcendental arguments only to perceivable subject matters. But this is unwarranted. There are no reasons why one cannot set up a transcendental argument to establish claims about unobservable entities or properties. A clear example of this is Sacks own discussion of the principle of causation according to which every event has a cause. In a nutshell the claim is that the irreversible order of our perceptions of two events can only be explained if the perceived states of affairs are put together in a such a way that one state could not 19 Maybe there is more to the notion of a situated thought that would enable it to accommodate such scenarios. But these need to be made more explicit. 20 If we consider a change as constituting an event and construe events in the standard fashion as the exemplification of a property by an object, does this claim not turn out to be analytically true? 21 Sacks, "The Nature,"

9 Skepticism and Varieties of Transcendental Argument happened except after the occurrence of the other. But this argument at most establishes an objective, irreversible, order between events themselves. As such, it is quite consistent with a Humean regularity view of causation. That the objective relation is also necessary is neither here nor there. Nomologically necessary relations are widely thought to be unperceivable. Having identified varieties of transcendental arguments and delineated their structure, I shall now turn to the question of their epistemic effectiveness. 2. The Epistemological Significance of Transcendental Arguments The epistemic potential of transcendental arguments is a function of both their type as well as their targets. So I begin with first-order transcendental arguments whose conclusions have first-order contents. Such arguments purport to show that the truth of certain statements is the necessary condition of the obtaining of certain types of states such as thoughts, experiences and so on. Let us start with the most well-known of such first-order transcendental arguments, namely, those used against Cartesian skepticism. Versions of this argument include Kant s argument in the Deduction and Strawson s variation on the same theme (the objectivity argument). I shall not directly challenge these arguments. Rather, I shall examine their legitimacy in the light of our construal of first-order transcendental arguments as species of IBE. Accordingly, what these arguments seek to show is that the realist hypothesis is a better explanation of our experiences and their specific features than other competing alternatives such as the Cartesian demon hypothesis, etc., in virtue of being simpler, more parsimonious and the like. As regards the structure of IBE, it was noted that the explanatory virtues, in terms of which we are supposed to discriminate between competing hypotheses, should be seen as surrogates for stating background assumptions about the subject matter under discussion. Accordingly, the success of IBE depends on the contexts where such an argument is being deployed for some of these contexts possess such a low degree of epistemic involvement that they do not allow invoking the required assumptions, thus, hindering the process of adjudicating between the competing hypotheses. The context of establishing the realist (external world) hypothesis is precisely such a context. Here the skeptic claims that our having the experiences we take to be of physical objects (with all their specific features) is compatible with the possibility of there being no such objects at all while a demon (or a superscientist) is inducing these experiences in us. No common grounds other than these experiences exist between the realist and the Cartesian skeptic. At a level this basic, we are, on pain of begging the question, precluded from appealing to background assumptions about the subject matter in question (be it the external world, the demon or whatever). 403

10 Hamid Vahid So if appeals to simplicity and other explanatory virtues are to be seen as involving background assumptions about a particular subject matter, IBE cannot be put to use to ground our preference for realism over, say, the Cartesian skeptical hypothesis. Now, if we take the view that first-order transcendental arguments are actually species of IBE, it follows that the corresponding anti-skeptical transcendental arguments cannot also be expected to deliver the goods. On the other hand, if the context of applying an IBE is one which possesses a high degree of epistemic involvement (as in scientific or ordinary contexts), then the invoking of the relevant background assumptions would not be unjustified because such contexts do not generally impose strict epistemic constraints on what can be said to be known. Accordingly, such inferences stand or fall depending on whether their pertinent background assumptions can be adequately defended. 22 The same holds for those transcendental arguments that collapse into such inferences. Let us now turn to second-order transcendental arguments whose conclusions are some belief statements. Stroud, we may recall, claims that it is not the truth of certain propositions that is needed to explain the obtaining of experience, knowledge, and the like, but the belief in those propositions. As such, typical transcendental arguments cannot undermine radical skeptical positions. Recently, however, Stroud has claimed that these modest transcendental arguments still possess anti-skeptical bite. A number of philosophers, who have taken the Stroudian line on the nature of transcendental arguments, have also tried to show that modest transcendental arguments possess significant anti-skeptical consequences. In what follows, I consider and reject two recent attempts in that direction before turning to Stroud s own argument. Using Kant s (anti-humean) response to the problem of induction as a case study, Ralph Walker has claimed that Kant s argument is not really intended to establish the reliability of induction but only that belief in such a hypothesis is indispensable if we are to have any thoughts at all. 23 He distinguishes between third-personal and second-personal transcendental arguments. The former aims at establishing certain truths about the world while the latter only aims to show that certain beliefs are indispensable for thought. So although Kant s transcendental argument for induction is of a second-personal variety, it is, he thinks, still effective against the skeptic as it justifies our inductive practices. I shall not dwell on Walker s contention about the typology of Kant s argument for induction, but only 22 For a defense of this position see Hamid Vahid, "Realism and the Epistemological Significance of Inference to the Best Explanation," Dialogue XL (2001): Walker, "Induction". 404

11 Skepticism and Varieties of Transcendental Argument concern myself with assessing his claim that it justifies our belief in induction. I think the claim is unwarranted. Although Walker admits that there may also be transcendental arguments which are third-personally valid, 24 he thinks they are generally problematic. In any case, Kant s argument for induction is not third-personal valid because he himself admits that he cannot show space and time to be the only possible forms of sensible intuition. But, taken second-personally, the argument, says Walker, is plausible because our experience has spatio-temporal character and so space and time must be a priori forms of intuition. As for non-spatio-temporal forms of sensible intuitions, Walker dismisses the possibility as irrelevant on the ground that [p]eople who claimed not to have spatio-temporal experience would not be worth arguing with, and if their experience really was different we should be unable to recognize them as having experience, or concepts, or language at all. 25 Walker s reasoning is, however, problematic as his latter remarks actually constitute a thirdpersonal transcendental argument (similar indeed to Davidson s well-known argument 26 against the possibility of radically different conceptual schemes). For what it says is that for people to be interpretable (i.e., being recognized as having experience, concepts, etc.), their experiences must be similar to ours. And since people are interpretable, the proposition people s experiences have spatio-temporal structure must be true. Thus, Walker s illustration of the problematic nature of third-personal transcendental arguments presupposes their validity. Let us, however, return to his claim about the anti-skeptical potentials of Kant s argument for the principle of induction. An early sign of confusion appears when Walker introduces a distinction in order to elucidate what the argument is intended to achieve: [R]efuting Hume s skepticism about induction would not be a matter of showing that our expectations will continue to be fulfilled. It would be a matter of showing that it is rational to expec that they will, not just as a matter of custom or habit. What Kant tries to do is to show by means of a transcendental argument that this is indeed rational. 27 However, if we take rational in an epistemic, truth conducive sense, then Walker s distinction is no longer tenable, for to show that belief in induction is justified (rational) is to show that it is more likely to be true. Walker further elaborates the sense in which transcendental arguments are, in his view, rational or justification-conferring. The skeptic calls into question 24 Walker, "Induction," Walker, "Induction," Donald Davidson, "On the Very Idea of a Conceptual Scheme," in his Inquiries into Truth and Interpretation (Oxford: Oxford University Press, 1984), Walker, "Induction,"

12 Hamid Vahid whether we are entitled or justified to rely on induction. Walker thinks that transcendental arguments, taken in the second-personal way, answer this. Certainly we are entitled to do so, and for the best possible reason: there is no serious alternative available to us. 28 He then considers the worry that secondpersonal transcendental arguments make no reference to truth, but goes on to say that third-personal transcendental arguments do no better. But Walker s having-no-other-option response hardly addresses the worry just raised. For what actually lies behind the objection is whether the type of justification purportedly provided by second-personal transcendental arguments has epistemic significance. To see what is at issue, we need to remind ourselves of various senses of justification (rationality). Justification is a goal-directed notion with different goals entering into different conceptions of justification. If one s goal is maximizing truth and minimizing falsity or achieving a comprehensive and accurate set of beliefs, then a belief serving this goal would be justified in an epistemic sense. But when our goal is, say, maximizing satisfaction of desire, then the ensuing concept of justification would be prudential. This gives rise to the possibility that one and the same belief might be justified for an agent in one sense but unjustified in another sense. For example, suppose it is rational to a high degree for a mediocre academic to believe that his believing that he is more talented than his colleagues will help him succeed at his job (by boosting his confidence). Then, all things being equal, it is practically rational for him to believe that he is more talented than his colleagues. But this very belief may be unjustified if viewed from the epistemic point of view of achieving an accurate belief system. In general, we may say that if an agent has a certain goal G and if, on careful reflection, he would believe that M is an effective (or indispensable) means to G, then, all else being equal, it is rational for him to bring about M. 29 We can now see that Walker s argument regarding the indispensability of belief in induction, as a means to the obtaining of thought and experience, at best confers justification on the belief in question only in a non-epistemic sense because the goal is not the maximization of truth but the obtaining of certain psychological states. Thus viewed, it is hardly encouraging news for the realist to be told that [s]econd-personal transcendental arguments are [his] only real defense against skeptical doubt Walker, "Induction," See, for example, Richard Foley, The Theory of Epistemic Rationality (Cambridge Mass.: Harvard University Press, 1987). Note that Walker s claim cannot be that having a justified belief (in a truth conducive sense) in induction, as opposed to merely believing it, is necessary for the possibility of thought, otherwise his second-personal argument would collapse into a third-personal transcendental argument. 30 Walker, "Induction,"

13 Skepticism and Varieties of Transcendental Argument A rather different tack has been tried by Robert Stern who also thinks that second-order transcendental arguments (which he calls belief-directed arguments) have anti-skeptical consequences. 31 He distinguishes between epistemic skepticism which requires certain knowledge in the face of skeptical doubts and justificatory skepticism which does not ask for conclusive reasons or justification but settles for much less, namely, reasons that are, by an agent s lights, justification-conferring. Stern thinks that this internal conception of justification is what Sosa has called reflective justification. Sosa identifies this species of justification with the outcome of the application of our deepest intellectual norms and a matter of perspectival coherence. 32 On this view, according to Stern, even the beliefs of the inhabitants of a demon world can be justified. He claims that Kant takes coherence to be one such norm so that if an agent s belief system is more coherent with the belief p than without it, then this belief is justified for that agent. 33 Second-order (belief-directed) transcendental arguments, says Stern, are justification-conferring precisely because they ensure that the adding of their conclusions to our belief system enhances its coherence. Stern s account raises a number of issues. One can see his appeal to the coherence norm as a way of supplementing Walker s proposal. The main problem with Walker s approach, as we saw, was that his underlying conception of justification didn t really hook up with truth resulting, at best, in a nonepistemically justified belief. One can see Stern s appeal to his version of the coherence theory of justification as a way of restoring the link between justification and truth. He admits, however, that he cannot establish that his coherence norm is truth conducive but also denies that such norms are thereby rubbed of their justificatory capacity, insofar as they seem to offer the best guide to truth from where we are. 34 But this remark hardly addresses the question of why his coherence-based conception of justification is truth conducive. And there are also the standard objections to the claim that mere internal coherence yields a reason for thinking that the beliefs of a system are true. These include the possibility of alternative, equally coherent systems; gaining epistemic access to the coherence of one s belief system (especially if coherence is not merely understood in a negative 31 Robert Stern, "On Kant s Response to Hume: The Second Analogy as Transcendental Argument," in Transcendental Arguments, ed. Robert Stern, See various articles in Ernest Sosa, Knowledge in Perspective (Cambridge: Cambridge University Press, 1991). 33 This requires that we be able to determine whether the coherence of our belief system has been enhanced, an ability whose psychological feasibility is doubtful. 34 Stern, "Kant s Response,"

14 Hamid Vahid sense but also as meaning that the beliefs in question must positively support one another). Moreover, it is not clear that the truth-conducivity of this particular brand of the coherence norm is compatible with what Stern claims about the justified status of the beliefs held by the inhabitants of a demon world for such beliefs are, by hypothesis, systematically false. Indeed, Sosa s views on this issue are more complex than are made to appear here. Sosa makes a distinction between aptness and justification. An apt belief is one that is produced by a reliable or virtuous faculty in the environment in which it is operating. A justified belief, on the other hand, is construed in terms of the notion of an epistemic perspective which is, in turn, cashed out as consisting of meta-beliefs concerning the faculty (responsible for producing the target belief) and its reliability. Commensurate with this distinction, Sosa makes another distinction between animal knowledge (which requires only apt belief) and reflective knowledge (requiring both apt and justified belief). Sosa is not ideally clear about the relationship between the notions of justification and aptness. But it seems that he takes these notions to be independent of one another. An apt belief is not automatically justified, and a justified belief need not be apt. This interpretation is further supported by Sosa's claim that while animal knowledge involves only apt belief, reflective knowledge requires both apt and justified belief. The fact that both justification and aptness are required for reflective knowledge shows that neither implies the other. And this is as it should be for Sosa regards justification [as] amount[ing] to a sort of inner coherence. 35 Sosa's response to the problem of the status of beliefs in demon scenarios thus involves appealing to the distinction between aptness and justification, but, in so doing, he relativizes assessment of the epistemic status of the victim's beliefs in the demon world: Relative to the demon s [environment] D, the victim s belief may be inapt and even unjustified Even so, relative to our environment the beliefs of the demon s victim may still be both apt and valuably justified. 36 Justificationrelativity does not seem to be congenial to Stern s concerns. So given the preceding problems for Stern s coherence theory of justification, it is not clear whether he would wish to follow Sosa s lead in defending the truthconductivity of his conception of justification. To avoid problems of exegesis, however, I shall propose a dilemma for Stern s account and argue that both horns of the dilemma undermine its viability. Let us begin by assuming that coherence is not truth conducive. In that case, Stern s account can hardly be seen as an improvement on Walker s indispensability argument. Both yield, at most, justified beliefs in a 35 Sosa, Knowledge in Perspective, Sosa, Knowledge in Perspective,

15 Skepticism and Varieties of Transcendental Argument non-epistemic sense that could hardly counter the threat of skepticism. Suppose however that Stern s brand of coherence theory is truth conducive assuming that the standard objections to such a theory, especially the problem of epistemic access, have been satisfactorily dealt with. Then, transcendental arguments are rendered redundant. For to see whether a belief p is justified, all we need to do is to find out, as Stern puts it, whether our belief-set is more coherent with the belief that p as a member than without it. 37 There will thus be no need to involve transcendental arguments. This creates a dilemma for Stern s account of transcendental arguments analogous to that posed by Stroud for the standard readings of such arguments. Either second-order transcendental arguments yield non-truth-conducive coherence in which case they lack anti skeptical potentials, or they generate truth conducive coherence in which case they will become redundant. 38 Let us now turn to Stroud s own gloss on the epistemic efficacy of the second-order transcendental arguments. Although he rejects ambitious first-order transcendental arguments, he thinks that the second-order variety still possesses some epistemic value in the face of the skeptical challenge, and illustrates this through the following second-order transcendental argument. 39 Suppose it has been established (as claimed by Strawson) that to think of a world independent of us we must think of it as containing enduring particular objects. What does this entail? Well, if the claim is true, it follows that we could not think of there being anything in the world (such as people) without thinking that there are enduring objects. This, in turn, entails that we could not think of these people as believing things and in particular as believing that there are enduring particulars without believing that there are enduring particulars. So we could not ascribe that belief to people without thinking that it is true. In other words, we cannot consistently find that people are wrong in believing that there are enduring particulars. This does not show that the belief in enduring objects is true but only that we cannot consistently deny it. As such the belief in question would be indispensable to our having any thoughts or beliefs about the world at all. Stroud takes this to imply that certain beliefs are invulnerable in the sense that we could never see ourselves as holding the beliefs in question and being mistaken. 40 Let us assume that Stroud is able to extend this treatment to all second-order transcendental arguments and draw the same conclusion, viz., that we could never 37 Stern, "Kant s Response," For further criticisms see Mark Sacks, "Transcendental Arguments and the Inference to Reality: A Reply to Stern," in Transcendental Arguments, Barry Stroud, "The Goal of Transcendental Arguments," in Transcendental Arguments, Stroud, "The Goal of Transcendental Arguments,"

16 Hamid Vahid see ourselves as holding the relevant beliefs and being mistaken. What does follow from this? The immediate upshot is to ensure the consistency of our belief system. Applying the relevant transcendental arguments, we are, in effect, implying that we can attribute certain beliefs to people (including ourselves) only if we already hold them to be true. This, at most, guarantees the consistency, not truth, of our beliefs. Indeed Stroud s argument is just an application of Davidson s general requirement on a correct theory of interpretation, viz., the principle of charity. According to Davidson to attribute thought and language to speakers we must assume that they hold similar beliefs as ours. Indeed, Stroud, himself goes on to illustrate his point by referring to the closely related position of Donald Davidson who, from conditions of belief ascription, or what he calls radical interpretation, concludes that most of our beliefs must be true. 41 Stroud, however, seems to think that Davidson holds the strong view that the application of charity guarantees the truth of most of our beliefs are true, and instead suggests the weaker conclusion that we cannot attribute beliefs to speakers which we see as mistaken. But this is consistent with Davidson s early formulations of charity which only require assigning truth conditions to alien sentences that make native speakers right when plausibly possible, according, of course, to our own view of what is right. 42 For all we know, however, the interpreter s beliefs might very well be mistaken. So Davidson s principle of charity at most guarantees agreement (consistency) between the interpreter and the speaker. This means that the epistemic efficacy of second-order transcendental arguments goes only as far as establishing the coherence of a belief-set. 43 The problem can be made more explicit once we consider Stroud s claim that beliefs borne out by second-order transcendental arguments are invulnerable. For invulnerability is a non-truth-conducive feature of such beliefs and, as such, of little consequence for defeating skepticism. 44 When Stroud says, of such beliefs, that we could not see ourselves as holding them and being mistaken, he is only highlighting the fact that we would fall into a pragmatic (not epistemic) trap if we deny them. This is very much like the position we get ourselves into when we assert the so-called Moorean s sentences of the type p is false but I believe that p, ~p & IBp, for to assert such a sentence is to assert ~p (i.e., to express the belief that p is mistaken) and yet go on to say that one believes that p. What is peculiar about 41 Stroud, "The Goal of Transcendental Arguments," Donald Davidson, "Radical Interpretation," Inquiries into Truth and Interpretation, For further elaboration of this point see Hamid Vahid, "Charity, Supervenience and Skepticism," Metaphilosophy 32, 3 (2001): It is for this reason that Davidson himself adds an externalist slant to the principle of charity to give it some epistemic bite. See Vahid, "Charity, Supervenience and Skepticism" for elaboration. 410

17 Skepticism and Varieties of Transcendental Argument Moorean sentences is that while they have coherent truth conditions, their assertion lands us in pragmatic, not logical, contradiction. Our inconsistency arises not from what we are claiming but from the fact that we are claiming it. This squares very nicely with the proposed role of the principle of charity in constituting the epistemic significance of second-order transcendental arguments. For once the principle of charity is recognized as being constitutive of intentional ascription, of what it is to be a speaker at all, we can see how the assertion of Moorean sentences undermines charity. Consider an instance of p & ~IBp, say, It is raining but I do not believe that it is raining. By asserting p, or assenting to p, one is performing a speech act to communicate certain information that p. Thus, when an agent assents to It is raining the default position is to interpret the utterance in such a way that it is true just in case it is raining, i.e., take it to mean it is raining. And, assuming Davidson s strictures on interpretation, to infer, in accordance with the principle of charity, that she believes that it is raining. So when, having assented to it is raining, the speaker goes on to assert that she does not believe that it is raining, this would be a clear case in which the principle of charity is undermined. So it is because the assertion of Moorean sentences contravenes the principle of charity that these sentences are defective. 45 Thus, when Stroud says that second-order transcendental arguments have epistemic value because one could not see himself as holding the belief in their conclusion p, say, There are enduring objects, and being mistaken, he is merely saying that the sentence ~p & IBp is Moorean in that its assertion is paradoxical. But the paradox is merely pragmatic and this, in turn, reflects the epistemic import of second-order transcendental arguments. As noted earlier, such arguments at best show that their conclusions are pragmatically, not epistemically, justified. As such they cannot be of much comfort to the anti-skeptic. To conclude, on any of the accounts discussed so far, transcendental arguments, of first-order or second-order variety, seem to lack significant epistemic bite. 45 See Hamid Vahid, "Radical Interpretation and Moore s Paradox," Theoria 74 (2008):